The macro is used by reading the macro into SAS with an "include" statement:



     %inc [[path and filename where macro is found]]



and then calling the macro by typing



     %lregmc1( parameter1=choice1, parameter2=choice2, ...)



where the parameters and choices are as follows:



DATA          SAS dataset to be used in fitting the model.  The default is the last dataset created.



YVAR          The variable that contains the imperfect diagnostic test result.  YVAR must be a 

                     binary variable which takes the values 1 or 0.



XVAR          The predictor variables in the model.  You have to create your own intercept 

                     variable if you want an intercept in the model.



SENS           The variable which provides the sensitivity of the diagnostic test which generated 

                     YVAR.  SENS must be a variable in the dataset and so it takes a value for each 

                     observation in the dataset.  In many applications, SENS will assume the same value 

                     for each observation.  The default is 1.



SPEC            Analogous to SENS, SPEC specifies the variable which provides the specificity of   

                     the diagnostic test.  The default is 1.



BETA           Initial estimates for the regression coefficients.  This is an optional parameter and 

                     may only be necessary if there is difficulty reaching convergency with the initial 

                     estimates chosen by the program.

     

ITER            The maximum number of iterations performed by the program.  The default number 

                     is 50.



CRIT            The convergence criteria.  The default value is .00001.



MISSDEL    This specified how missing values are handled.  If "yes" then all observations which 

                     have missing values for either YVAR, XVAR, SENS, or SPEC will be 

                     automatically removed from the dataset before the program model is fit.  The 

                     default is "yes".  I cannot think of any reason not to specify MISSDEL="yes".



PRINTER     This specifies whether the results from each iteration are printed.  The iteration 

                     results will not be printed unless PRINTER="yes".



=====================================================================



Example:


The following program reads in some data and fits a simple model under the assumption that the

sensitivity of the diagnostic test is .8 and the specificity is .9:



%inc "c:lregmc1.sas";


data a;

input exposures testres;

sensvar-.8;

specvar=.9;

intcept=1;

cards;

1 1

1 1

1 1

1 0

1 0

0 1

0 1

0 0

0 0

0 0
;



%lregmc1 (data=a,yvar=testres,xvar=intcept exposure, sens=sensvar,spec=specvar);

run;



This produces the following output:

                   The SAS System           09:50 Friday, August 16, 1996


Logistic Regression when outcome is measure with error and sensitivity and specificity are known



Data File: a


Outcome variable: TESTRES

Predictors: INTCEPT EXPOSURE

Sensitivity variable: SENSVAR

Specificity variable: SPECVAR


Number of non-missing obs used:     10



Averages of Outcome variable and Predictors:


TESTRES     INTCEPT     EXPOSURE     SENSVAR     SPECVAR

 0.5          1.0           0.5         0.8        0.9


Initial estimate of regression coefficients:

INTCEPT          0.4

EXPOSURE         0.2


     NOTE: Convergence after 19 iteration(s).


Variance/covariance of estimates:


                    INTCEPT        EXPOSURE

INTCEPT           1.633333        -1.633333

EXPOSURE         -1.633333         3.9853066 



Estimate with s.e., z-score, p-value, and confidence limits:

           Estimate       SE            z           pvalue        OR       lower         upper

INTCEPT   -0.287681     1.2780     -0.2251        0.8219       0.75        0.06           9.18

EXPOSURE  1.2039607     1.9963      0.6031        0.5464       3.33        0.07       166.79


Log-likelihood=-6.730117